1 http://jbiol.com/content/2/4/27 2 3 A functional genomic analysis of cell 4 5 morphology using RNA interference 6 7 AA Kiger , B Baum, S Jones, MR Jones, A Coulson, 8 9 C Echeverri and N Perrimon 10 11 Background 12 13 The morphological diversity of animal cells 14 results largely from differences in the 15 lineage-specific expression and control of 16 cytoskeletal regulators. Cells in culture have 17 been widely used to characterize morphogenetic 18 events, for example the dynamics and organization 19 of filamentous actin and microtubules in adherent 20 and motile cells. Few metazoan cell systems, 21 however, permit the use of genetic analysis to 22 identify the complement of genes contributing 23 to the generation of cell shape. 24 25 RNA interference (RNAi) has revolutionized 26 the functional analysis of genes identified 27 by genomic sequencing [1-3]. Several factors 28 make RNAi in Drosophila cell cultures an 29 excellent approach for such functional 30 genomic analysis of animal cell form. 31 The availability of well-annotated Drosophila 32 genomic sequence simplifies the design of 33 gene-specific double-stranded RNAs (dsRNAs) 34 [4]. Furthermore, the Drosophila genome encodes 35 homologs of over 60% of human disease genes 36 [5] and lacks some of the genetic redundancy 37 observed in vertebrates. RNAi in Drosophila 38 cells is efficient, reducing or eliminating 39 target-gene expression to elicit partial to 40 complete loss-of-function phenotypes upon the 41 simple addition of dsRNA to the culture medium 42 [6]. Finally, the well-established genetic 43 techniques for Drosophila allow comparisons to 44 be made between loss-of-function cell-culture 45 phenotypes and those observed in tissues 46 of corresponding mutant flies. 47 48 In order to develop a cell-based approach 49 for the study of gene functions involved in 50 morphogenesis, we developed a high-throughput 51 RNAi screening methodology in Drosophila cell 52 cultures that is applicable to the study of 53 a wide range of cellular behaviors (Figure 54 1a). This approach involves the following 55 steps: first, the design and synthesis of a 56 gene-specific dsRNA library; second, incubation 57 of Drosophila cells with the dsRNAs in 384-well 58 assay plates (in serum-free medium or with 59 transfection reagents, depending on the cell 60 line); and third, optional induction of a cell 61 behavior, followed by detection of luminescent 62 or fluorescent signals using a plate reader or 63 an automated microscope. 64 65 Here, we describe the establishment of an RNAi 66 functional approach applied to the study of cell 67 morphology. Using images acquired by automated 68 microscopy, we visualized phenotypic changes 69 resulting from reverse-functional analysis by the 70 treatment of Drosophila cells in culture with 71 gene-specific dsRNAs. We were able to observe 72 and characterize a wide range of phenotypes 73 affecting cytoskeletal organization and cell 74 shape, and from these, to identify sets 75 of genes required for distinct round versus flat 76 cell morphologies. 77 78 Results and discussion 79 80 Drosophila cell morphology in cultures 81 82 We began by surveying existing Drosophila cell 83 lines to identify those with distinct 84 but uniform cell shape, size and adhesion 85 properties. For a comparative study, we chose 86 to further characterize two well-established 87 lines, Kc167 and S2R+ cells [7-9], because 88 of their differences in cell shape. 89 Although both lines apparently derived from 90 embryonic hemocytes (blood cells), Kc167 cells 91 are small and round (10 mm; Figure 1b), whereas 92 S2R+ cells are large, flat and strongly adherent 93 to glass, plastic and extracellular matrix 94 (averaging 50 mm; Figure 1c). The stereotypical 95 morphology of each cell line could be modified 96 in specific ways using drugs that perturb 97 cytoskeletal function (for example cytochalasin, 98 latrunculin, nocodazole or colchicine; see Figure 99 1d), ecdysone hormone treatment (Figure 1e), 100 substrate-induced cell polarization (phagocytosis 101 of bacteria or polystyrene beads; data not shown) 102 or gene-specific RNAi (Figure 1a). For example, 103 treatment with a drug that prevents the 104 polymerization of filamentous (F-) actin caused 105 Kc167 cells to develop long microtubule-rich 106 processes, a morphological change similar to 107 that observed upon treatment with dsRNA 108 corresponding to the gene encoding Cdc42 GTPase. 109 Thus, both cell types could be used with RNAi 110 to assay single-gene functions that contribute 111 to cytoskeletal organization and cell shape. 112 113 RNAi assay for cell morphology phenotypes 114 115 We set out to conduct parallel RNAi screens with 116 a microscopy-based visual assay to identify genes 117 required for the characteristic round versus flat 118 morphology of Kc167 and S2R+ cells, respectively 119 (Figure 1a,b,c). By labeling actin filaments, 120 microtubules and DNA, it was possible to assay 121 a wide range of cellular behaviors in these cell 122 types, including cytoskeletal organization, cell 123 shape, cell growth, cell-cycle progression, 124 cytokinesis, substrate adhesion and cell 125 viability. 126 127 We used dsRNA to Rho1, a gene required for 128 cytokinesis [10], to optimize conditions for 129 RNAi in a 384-well plate format. The addition 130 of 0.3 mg Rho1 dsRNA to cells for a minimum 131 of 3 days in culture generated a penetrant 132 multinucleated cell phenotype (62 100% per 133 imaged field over five wells). Under these 134 conditions, RNAi was effective in both cell 135 types, as judged by the appearance of phenotypes 136 and/or depletion of the targeted gene products. 137 When screening many genes under a single assay 138 condition, several factors could influence the 139 efficiency of RNAi. Given that dsRNA targets 140 the destruction of endogenous mRNA, the efficacy 141 of RNAi and thus the phenotypic strength could 142 reflect gene- and cell-type-specific differences 143 in mRNA levels, the levels and stability of the 144 preexisting protein pool and/or the potency of 145 the chosen dsRNA targeting sequence. In one 146 example, a longer RNAi incubation time of 5 147 days was necessary to completely deplete the 148 Capulet/Cyclase associated protein, as detected 149 by western blot (although phenotypes affecting 150 F-actin organization were observed by 3 days; 151 data not shown). Thus, it is assumed that the 152 strength or penetrance of RNAi-induced phenotypes 153 observed under one screening condition could 154 vary marginally for any specific gene target or 155 cell type. We reasoned that screening under 156 'hypomorphic' conditions has the advantage of 157 enabling the effects of gene product depletion 158 to be analyzed rather than its 159 terminal consequences (that is, potential 160 cell lethality). Finally, differences in the 161 phenotypic effects of targeting the same gene 162 with RNAi in two different cell types could 163 reflect true cell-type differences in the 164 function of the targeted genes. 165 166 Selection and generation of gene-specific dsRNAs 167 168 Screens of RNAi morphological phenotypes required 169 the generation of a dsRNA library. In order to 170 allow an assessment of the overall success of 171 such an RNAi screening approach in Drosophila 172 cells, we generated a selected set of 1,042 173 dsRNAs targeting 994 different genes. The set 174 of genes represented in the library was chosen on 175 the basis of primary sequence to include the vast 176 majority of those predicted to encode 177 signaling components and cytoskeletal regulators 178 that could affect diverse cellular processes 179 (a complete list of the selected categories of 180 predicted gene functions are listed in Table 181 1; all targeted genes and primer sequences are 182 listed in Additional data file 1). Gene-specific 183 dsRNAs averaging 800 base pairs (bp) in length 184 were generated by in vitro transcription, using 185 selectively amplified products from Drosophila 186 genomic DNA as templates, then aliquoted into 187 384-well optical bottom plates for image-based 188 screens (see the Materials and methods section). 189 190 The dsRNA collection was selected to enrich 191 for genes encoding classes of central cell 192 regulators, including putative GTPases, GTPase 193 regulators, kinases and phosphatases that can act 194 together as part of signaling pathways to control 195 diverse cellular processes. We also selected 196 cytoskeletal proteins and cell-cycle regulators 197 predicted to be expressed and required in 198 most cells. We favored target selection on the 199 basis of identifiable domains within the primary 200 sequence in order to enrich for both functionally 201 known and uncharacterized genes affecting a 202 wide range of processes. Choosing genes from 203 one chromosomal region would be likely to yield 204 fewer visible phenotypes, whereas choosing genes 205 on the basis of their expression in 206 existing cell lines would assume a correlation 207 between expression levels and function. 208